open-claudecode-research

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill includes literal API-key-like strings (e.g., "sk-ant-your-key-here") and shows commands/files that embed or print those tokens (settings.json, export, echo), which would require an LLM to handle or reproduce secret values verbatim.

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). The GitHub repo is an unknown third‑party clone/distribution of a runnable CLI (potentially containing native binaries) and the placeholder third‑party proxy domain is an arbitrary endpoint that could be used to capture tokens — both are untrusted sources for downloading/executing software.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs cloning and then running code from the remote repository https://github.com/LING71671/Open-ClaudeCode.git (e.g., git clone ... then node Open-ClaudeCode/package/cli.js in CI), which fetches remote code that is executed at runtime and can control agent prompts.